Art And Technology Integration: CNN-Based Image Style Transfer Research
DOI:
https://doi.org/10.54097/3th7y626Keywords:
Neural Style Transfer; Convolutional Neural Networks; Semantic Segmentation.Abstract
This research presents a CNN-based image style transfer framework that integrates layered feature extraction with region-aware stylization. Traditional neural style transfer methods often apply styles globally across an image, blurring boundaries between subjects and backgrounds and reducing fine-grained stylistic control. To address this limitation, the proposed method combines multi-layer feature modeling using the VGG19 network with semantic segmentation from DeepLabv3 to distinguish subject and background regions. By separating feature maps using binary masks, the system computes independent style losses for each region while maintaining the overall content structure via a content loss function. Experimental results on a dataset of diverse images—including portraits, architecture, and landscapes—demonstrate that the framework achieves clearer structural preservation and more balanced stylization than conventional global methods. Parameter studies further show that adjusting the weights of content, subject style, and background style enables flexible artistic control. The approach offers potential applications in digital art creation, cultural heritage visualization, and AI-assisted design.
Downloads
References
[1] Image Style Transfer, the earliest formalization inspired by Gatys et al. (2015).
[2] Gatys, L. A., Ecker, A. S., & Bethge, M. (2015). A Neural Algorithm of Artistic Style. arXiv:1508.06576.
[3] LeCun, Y., Bottou, L., Bengio, Y., & Haffner, P. (1998). Gradient-Based Learning Applied to Document Recognition (LeNet). Proceedings of the IEEE.
[4] Simonyan, K., & Zisserman, A. (2014). Very Deep Convolutional Networks for Large-Scale Image Recognition (VGG). arXiv:1409.1556.
[5] Chen, L. C., Papandreou, G., Kokkinos, I., Murphy, K., & Yuille, A. L. (2017). DeepLabv3: Semantic Image Segmentation. arXiv:1706.05587.
[6] Johnson, J., Alahi, A., & Fei-Fei, L. (2016). Perceptual Losses for Real-Time Style Transfer and Super-Resolution. ECCV.
[7] Ulyanov, D., Vedaldi, A., & Lempitsky, V. (2016). Instance Normalization: The Missing Ingredient for Fast Stylization. arXiv:1607.08022.
[8] Zhu, J. Y., Park, T., Isola, P., & Efros, A. A. (2017). Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks. ICCV.
[9] Karras, T., Laine, S., & Aila, T. (2019). A Style-Based Generator Architecture for GANs. CVPR.
[10] Li, C., & Wand, M. (2016). Combining Markov Random Fields and CNNs for Image Synthesis. CVPR.
[11] Chen, Y., et al. (2017). Semantic Style Transfer and Turning Two-Bit Doodles into Fine Artworks. arXiv.
[12] Dumoulin, V., Shlens, J., & Kudlur, M. (2017). Learned Representations for Artistic Style. ICLR.
[13] Huang, X., & Belongie, S. (2017). Arbitrary Style Transfer in Real-Time with Adaptive Instance Normalization. ICCV.
[14] Zhang, H., Goodfellow, I., et al. (2019). Self-Attention Generative Adversarial Networks. ICML.
[15] Sanakoyeu, A., Kotovenko, D., et al. (2018). A Style-Aware Content Loss for Real-Time HD Style Transfer. ECCV.
[16] Yao, A., et al. (2018). Dual Discriminator GANs. NeurIPS.
[17] Li, C., & Wand, M. (2016). Precomputed Real-Time Texture Synthesis with Markovian GANs. ECCV.
[18] Mechrez, R., Talmi, I., & Zelnik-Manor, L. (2018). The Contextual Loss for Image Transformation with Non-Aligned Data. ECCV.
[19] Park, J., & Lee, K. M. (2019). Style-Aware GANs for Realistic Image Synthesis. CVPR.
[20] Kotovenko, D., Sanakoyeu, A., et al. (2019). Content and Style Disentanglement for Artistic Style Transfer. ICCV.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Highlights in Science, Engineering and Technology
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
